S12XEP signed 32 substraction issue

[mroczeks]

typedef signed long int S32;S32 b;S32 x1;S32 x2;

......

 b = x2 - x1;

As a result only lower 16 bits are present in b variable. Higher 16 bits are lost somewhere. I am observing it under CodeWarrior debugger (HiWave). Why?

[kef]

It should be fine. 

It is not clear what is storage class of b,x1 and x2. It is not clear what CW product version you are using? CW 4.6? Which compiler, C++ for XGATE?

Did you try disassembling C code? Does it look right? Could you show disassembled code  for b = x2 - x1? (Use search to find commented C source line in disassembled code).